Principal Software Engineer

Global X ETFs•New York, NY

1d•$230,000 - $245,000•Hybrid

About The Position

We are hiring a founding engineer for the firm's proprietary investment platform — in effect, the operating system for our trading and portfolio management desks. Global X operates a large and growing ETF platform across multiple markets. Our investment desk runs on proprietary technology built in-house for portfolio management, trading, and the operational workflow around them, and we are investing significantly in advancing it as the product range and the number of markets grow. The engineering problems here are genuinely interesting ones. Real-time market data under back-pressure. Sub-second recalculation across a continuously changing book. Idempotent posting to external systems of record, where a duplicate or dropped message has direct financial consequences. Deterministic replay from an immutable event log. Multiple markets, calendars, and time zones. And a correctness standard set by a regulated business rather than by convenience. This role will own services end to end — design, implementation, deployment, and production — across a Java service tier and a Python quantitative tier on AWS. The decisions you make here shape the platform for years and set the engineering standard for the team that grows around it. Agentic coding tools are core to how we work, and AI is core to what we are building: the platform is designed so that agents are first-class participants in it rather than a capability layered on afterward. If you have been looking for a role where the scope is wide, the problems are hard, and the tooling keeps up with your ambition, this is that role.

Requirements

8+ years building and operating production backend services, including ownership of systems where failure had real consequences.
Deep experience with distributed, fault-tolerant systems: event-driven architectures, idempotency and delivery guarantees, back-pressure, graceful degradation, and the failure modes that only appear in production.
Expert Java (modern LTS releases, Spring Boot) and strong production Python, and comfortable moving between the two rather than specializing in one.
Agentic coding tools (Claude Code, Codex, Cursor, or equivalent) as part of your daily production work. We are interested in how you use them — where they accelerate you, where they mislead, and how you verify.
Cloud-native engineering on AWS: containerized compute, relational data, caching, queueing and eventing, and infrastructure-as-code.
Experience in a domain where correctness was non-negotiable — financial calculation, accounting, billing, or similar — with the testing discipline that implies.
A track record of owning systems end to end and of setting technical direction others follow.
Strong written communication. Specifications and design documents are load-bearing here, for the humans and for the tooling.

Nice To Haves

Buy-side or asset management domain: positions and holdings, transactions, corporate actions, benchmarks and index data, NAV, and the order lifecycle. ETF or 1940-Act fund experience is a strong plus.
Real-time market data and execution vendor integration (Bloomberg, FactSet, or equivalents); FIX protocol experience.
Building or consuming MCP servers, tool interfaces, or other agent-facing surfaces over production systems.
Spec-driven development workflows and maintaining agent steering documentation across a multi-service codebase.
Cross-language service interop; cloud data warehouse and distributed cache technologies.
Engineering in a regulated environment: SEC record keeping (Rule 204-2), Regulation S-P, SOC 1 controls, and the evidentiary questions raised by AI-assisted change.

Responsibilities

Design, build, and operate containerized services on AWS across a Java service tier and a Python analytical tier, owning them in production and in development.
Build for continuous operation through partial failure. A market data feed drops, a counterparty is unreachable, a batch arrives late — the platform degrades in defined, observable ways rather than stopping, because the desk depends on it every trading day.
Design idempotent, exactly-once semantics wherever the platform posts to external systems of record, and keep the real-time calculation path fast under continuous change, with heavier analytical work off the critical path.
Implement the event-sourced audit trail and immutability guarantees that let any calculated value or user action be reconstructed exactly, and make deterministic replay a system property rather than an aspiration.
Shape APIs around the workflows users actually perform, with step-wise state, reversibility, and undo as first-class capabilities.
Build for multiple markets, calendars, and time zones. The desks this platform serves do not all trade the same hours, and the design should not assume they do.
Build the platform so AI agents are first-class participants — structured context, explicit action boundaries, and provenance on every state change, treated as cross-cutting infrastructure alongside logging and observability rather than as a feature.
Use agentic coding tools as a primary instrument of delivery. We scope ambitiously because of them, and we would rather give a strong engineer more surface area than more process.
Own the harness as well as the output: steering documents, specifications, test fixtures, and evaluation criteria that make agents genuinely productive on this codebase rather than generically capable.
Raise the review bar to match. When generation is cheap, verification is where craft lives, and correctness, security, and maintainability remain yours.
Track the tooling frontier and tell us when it moves. Work that took a week last quarter may not this quarter, and we want a team that notices and re-plans rather than continuing out of habit.
Work directly with portfolio managers, traders, operations, and compliance to understand the investment workflow in depth, then translate it into service boundaries and API contracts.
Hold the line on numerical correctness, with validation rigorous enough that results can be relied on without a second opinion.
Partner with the desk through each stage of delivery: investigate differences, separate genuine defects from expected variation, and drive them to resolution.
Set the standard for testing, CI/CD, observability, and secrets hygiene in a regulated environment where audit evidence is a deliverable in its own right.
Contribute architecture decision records, and leave a codebase legible to both the engineers and the agents that work on it next.
Help shape how the engineering function scales — hiring, standards, mentorship, and the operating model — as the team expands around the platform.